Skip to content

Add AI skill to run Jovian upgrade#11707

Draft
Mc01 wants to merge 2 commits intomc01/opcm-jovianfrom
mc01/jovian-skill
Draft

Add AI skill to run Jovian upgrade#11707
Mc01 wants to merge 2 commits intomc01/opcm-jovianfrom
mc01/jovian-skill

Conversation

@Mc01
Copy link
Copy Markdown
Contributor

@Mc01 Mc01 commented Apr 8, 2026

Description

Introduces a comprehensive AI skill (run-jovian-upgrade) that interactively guides engineers through the complete Jovian upgrade pipeline (v4.1.0 → v5.0.0 → Succinct v2) on a local Anvil fork. The skill orchestrates work across 6 repositories (Celo, Optimism, Optimism2, SuperchainOps, CeloSuperchainOps, Succinct) with checkpoint-based state verification between each upgrade phase.

Previously, running the Jovian upgrade locally required deep tribal knowledge of the correct script order, env vars, Safe nonce expectations, branch requirements, and cross-repo dependencies. This skill captures that knowledge in a structured, repeatable workflow that any engineer can follow — even without prior upgrade experience.

The skill covers: fork setup with port conflict handling, Safe multisig mocking, op-deployer bootstrap (v4/v5), Succinct FDG deployment with config-driven vkey population, SuperchainOps simulation, CeloSuperchainOps EIP-712 signing, sequential upgrade execution with 4 on-chain checkpoints, and domain-specific validations (new CeloSuperchainConfig, new ASR, game→ASR pointers, SystemConfig ownership transfer).

Additionally migrates the project knowledge base from the Claude Code-specific CLAUDE.md to a provider-agnostic .agent/AGENT.md structure, and consolidates all skills under .agent/skills/.

Other changes

  • Migrated existing skills (celo-release, node-cache-update) from .cursor/skills/ to .agent/skills/ (content unchanged)
  • Removed .agent/skills symlink, replaced with real directory
  • Removed .cursor/skills/ directory (content migrated)
  • CLAUDE.md replaced with a thin pointer to .agent/AGENT.md — keeps Claude Code compatibility while being LLM-agnostic
  • .agent/AGENT.md includes new sections: Repository Dependency Map, Network Configuration Reference (Safe addresses, proxy addresses, fork block numbers), Default Mocked Accounts, X-Ray Health Rules reference, and Upgrade Pipeline Timeline

Tested

Manually tested the full skill flow end-to-end on a local Anvil fork:

  • Phase 0: Validated all 6 repos (branches, binaries, artifacts, tools)
  • Phase 1: Background Anvil fork + mock-mainnet.sh (all 3 Safes validated)
  • Phase 2: Bootstrap v4 (OPCM deployed), Bootstrap v5 (OPCM deployed), Succinct v2 game impl deployed
  • Phase 4: Signed all 12 signatures (v4/v5/succ-v2 × clabs/council) via CeloSuperchainOps
  • Phase 5: Executed v4 → CP1 verified (v4.1.0 green), executed v5 → CP2 verified (v5.0.0 magenta, PermissionedGame 1.8.0), executed succ-v2 → CP3 verified (OPSuccinctGame updated, SystemConfig owner transferred to cLabs Safe)

Related issues

Backwards compatibility

No backwards compatibility concerns. This PR only adds new files under .agent/ and modifies CLAUDE.md to point to the new location. No existing scripts, contracts, or CI workflows are affected. The migrated skills retain identical content.

Documentation

  • .agent/AGENT.md — Unified project knowledge base (replaces CLAUDE.md content + Jovian ecosystem documentation)
  • .agent/skills/run-jovian-upgrade/SKILL.md — Interactive upgrade skill (1045 lines, 6 phases, 4 checkpoints)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant